Google and NASA are building a robot to help astronauts with medical emergencies

A robot helper, not a robot doctor

Google is working with NASA on a robotic assistant meant to help astronauts deal with health problems on long trips far from Earth, as Yahoo Tech reported. The robot would not act as a doctor. Its job is to support whoever on the crew serves as medical officer, and to take on other chores when nobody needs medical help.

Those chores include maintenance, moving equipment and managing supplies. The logic is that a small crew has little spare time, so a machine that can handle cargo and logistics, then switch over to medical support, makes better use of everyone's day.

The work builds on an earlier project called the Crew Medical Officer Digital Assistant. That is an AI system, which Google and NASA have been evaluating since 2025, designed to help astronauts size up a health problem when talking to doctors on Earth is slow or impractical. The robot adds a physical body to that software idea.

Why distance changes the problem

The driving issue is delay. Jim Kelly, Google's vice president for the public sector, pointed out that on a Mars mission a message and its reply could take up to 44 minutes round trip. In an emergency, waiting that long for instructions from mission control is a serious handicap, so more of the help has to be on board.

The source offers one worked example. Imagine an astronaut with intense back pain that might come from kidney stones. The astronaut could ask for an ultrasound, a scan that uses sound waves to produce pictures of the inside of the body. The robot would retrieve and prepare the equipment, then help position it.

Then the human takes over. The medical officer studies the images and decides what to do next. Diagnosis and treatment stay under human supervision, while the robot handles the fetching and setup.

Kelly described the long-term goal as a robot that follows simple spoken instructions while astronauts watch and direct it. Eventually it might help with routine health monitoring, such as capturing basic vitals or simple ultrasounds, always under the strict guidance of the crew medical officer.

What the 85% figure does and does not show

So far the evidence is thin. Kelly said that in the first laboratory trials the robot succeeded 85% of the time when asked to find and move a medical bag. For a basic task that is a reasonable start, but it is only that: a bag retrieval.

The result does not show that the robot can carry out a medical examination or cope with a complicated emergency. The source also does not say what success rate a real emergency would demand, so there is no yardstick for judging whether 85% is close to good enough.

Two pieces of technology support the effort. Google is adapting navigation software developed at NASA's Jet Propulsion Laboratory so the robot can move around and locate equipment. The team has also used MuJoCo, a physics simulator from DeepMind, which is a virtual world that models how objects move and collide. It lets them test movement and object handling before attempting more demanding jobs.

A proof of concept with few details

Google describes the project as a proof of concept. Early laboratory demonstrations are finished, but no fully integrated, space-ready robot has been confirmed. Several groups are involved: Google Cloud, NASA's Human Research Program, the robotics team at Johnson Space Center, and the Jet Propulsion Laboratory.

NASA itself has said little. According to the report, the agency declined to give detailed information, saying only that it is looking at several advanced AI initiatives with partners for possible future missions. No schedule, budget or flight plan appears in the coverage.

That makes this a story about direction rather than delivery. Still, the design problem is worth understanding well beyond space travel: build a tool that must work with slow contact to experts, high stakes and a human who makes the final call. The same question applies to any health AI you might meet on Earth: who checks its answers?